When roots license and when they respect semantico-syntactic structure in verbs Antje Roßdeutscher Universit¨ at Stuttgart Institut f¨ ur maschinelle Sprachverarbeitung – Logik und Sprachphilosophie – Azenbergstr. 12 D 70174 Stuttgart [email protected]Roots. Word formation from the perspective of ”core lexical elements”. Stuttgart 2009 2009.06.11 Roßdeutscher (IMS Stuttgart) Roots license, roots respect structure 2009.06.11 1 / 25
41
Embed
When roots license and when they respect semantico ... fileIn verbal constructions all roots serve to specify the functional v-head in a nunber of di erent ways on many routes. The
This document is posted to help you gain knowledge. Please leave a comment to let me know what you think about it! Share it to your friends and learn new things together.
Transcript
When roots license and when they respectsemantico-syntactic structure in verbs
IntroductionIn this talk I will present a way of constructing DRT-based semantic representations ofverbs from specifications of their roots.
• I assume three basic types of roots.(a) event type denoting roots
√arbeit (work),
√steig (rise),
√tauch (dive)
(b) property denoting roots and√
full (full),√
leer (empty),√
schließ (close)(c) sortal roots: material objects, e.g.
√deck (cover),
√lad (load); spatial regions,
e.g.√
ort (location) configurations; e.g.√
stapel pile; laws, e.g.√
regel (rule), etc..
• The sort of entity types denoted by a root makes it suitable for selection byv(erbal), a(djectival), or n(ominal) functional heads. They may alos enter otherconfigurations.
• In verbal constructions all roots serve to specify the functional v-head in a nunberof different ways on many routes. The three types of roots either license or respectstructure, in particular argument structure. In the first part of the talk I will showhow they do this.
• A restricted set of roots, with appropriate encyclopaedic properties, may mergewith the functional verbal head directly [Embick(2004)], leading to the syntacticand semantic structures of unergative and non-core-transitive verbs, [Levin(1999)],[Marantz(2005)].
IntroductionIn this talk I will present a way of constructing DRT-based semantic representations ofverbs from specifications of their roots.
• I assume three basic types of roots.(a) event type denoting roots
√arbeit (work),
√steig (rise),
√tauch (dive)
(b) property denoting roots and√
full (full),√
leer (empty),√
schließ (close)(c) sortal roots: material objects, e.g.
√deck (cover),
√lad (load); spatial regions,
e.g.√
ort (location) configurations; e.g.√
stapel pile; laws, e.g.√
regel (rule), etc..
• The sort of entity types denoted by a root makes it suitable for selection byv(erbal), a(djectival), or n(ominal) functional heads. They may alos enter otherconfigurations.
• In verbal constructions all roots serve to specify the functional v-head in a nunberof different ways on many routes. The three types of roots either license or respectstructure, in particular argument structure. In the first part of the talk I will showhow they do this.
• A restricted set of roots, with appropriate encyclopaedic properties, may mergewith the functional verbal head directly [Embick(2004)], leading to the syntacticand semantic structures of unergative and non-core-transitive verbs, [Levin(1999)],[Marantz(2005)].
IntroductionIn this talk I will present a way of constructing DRT-based semantic representations ofverbs from specifications of their roots.
• I assume three basic types of roots.(a) event type denoting roots
√arbeit (work),
√steig (rise),
√tauch (dive)
(b) property denoting roots and√
full (full),√
leer (empty),√
schließ (close)(c) sortal roots: material objects, e.g.
√deck (cover),
√lad (load); spatial regions,
e.g.√
ort (location) configurations; e.g.√
stapel pile; laws, e.g.√
regel (rule), etc..
• The sort of entity types denoted by a root makes it suitable for selection byv(erbal), a(djectival), or n(ominal) functional heads. They may alos enter otherconfigurations.
• In verbal constructions all roots serve to specify the functional v-head in a nunberof different ways on many routes. The three types of roots either license or respectstructure, in particular argument structure. In the first part of the talk I will showhow they do this.
• A restricted set of roots, with appropriate encyclopaedic properties, may mergewith the functional verbal head directly [Embick(2004)], leading to the syntacticand semantic structures of unergative and non-core-transitive verbs, [Levin(1999)],[Marantz(2005)].
IntroductionIn this talk I will present a way of constructing DRT-based semantic representations ofverbs from specifications of their roots.
• I assume three basic types of roots.(a) event type denoting roots
√arbeit (work),
√steig (rise),
√tauch (dive)
(b) property denoting roots and√
full (full),√
leer (empty),√
schließ (close)(c) sortal roots: material objects, e.g.
√deck (cover),
√lad (load); spatial regions,
e.g.√
ort (location) configurations; e.g.√
stapel pile; laws, e.g.√
regel (rule), etc..
• The sort of entity types denoted by a root makes it suitable for selection byv(erbal), a(djectival), or n(ominal) functional heads. They may alos enter otherconfigurations.
• In verbal constructions all roots serve to specify the functional v-head in a nunberof different ways on many routes. The three types of roots either license or respectstructure, in particular argument structure. In the first part of the talk I will showhow they do this.
• A restricted set of roots, with appropriate encyclopaedic properties, may mergewith the functional verbal head directly [Embick(2004)], leading to the syntacticand semantic structures of unergative and non-core-transitive verbs, [Levin(1999)],[Marantz(2005)].
IntroductionIn this talk I will present a way of constructing DRT-based semantic representations ofverbs from specifications of their roots.
• I assume three basic types of roots.(a) event type denoting roots
√arbeit (work),
√steig (rise),
√tauch (dive)
(b) property denoting roots and√
full (full),√
leer (empty),√
schließ (close)(c) sortal roots: material objects, e.g.
√deck (cover),
√lad (load); spatial regions,
e.g.√
ort (location) configurations; e.g.√
stapel pile; laws, e.g.√
regel (rule), etc..
• The sort of entity types denoted by a root makes it suitable for selection byv(erbal), a(djectival), or n(ominal) functional heads. They may alos enter otherconfigurations.
• In verbal constructions all roots serve to specify the functional v-head in a nunberof different ways on many routes. The three types of roots either license or respectstructure, in particular argument structure. In the first part of the talk I will showhow they do this.
• A restricted set of roots, with appropriate encyclopaedic properties, may mergewith the functional verbal head directly [Embick(2004)], leading to the syntacticand semantic structures of unergative and non-core-transitive verbs, [Levin(1999)],[Marantz(2005)].
Morpho-phonological form via move Interpretation via merge
Head Movement Constraint[Baker(1988)]An X0 can only move into an Y0
which properly governs it
(i) v introduces a binding conditionfor the referential argument e’(ii) e’ is a processthe root’s event predicate applies toe’: e’ is substituted for e.(iii) vP external subject[Kratzer(1996)]
Morpho-phonological form via move Interpretation via merge
Head Movement Constraint[Baker(1988)]An X0 can only move into an Y0
which properly governs it
(i) v introduces a binding conditionfor the referential argument e’(ii) e’ is a processthe root’s event predicate applies toe’: e’ is substituted for e.(iii) vP external subject[Kratzer(1996)]
property rootsJohnny Depp fullte die Kinokassen (mit Geld) die Fullung der Kinokassen’ J.D. filled the cinemas’ tills (with money)’
(i)rP
�� HHKinokassen
√full
(ii)v
���
HHH
rP
�� HHKinokassen t1
v
��HH√full1 v
vP
*s0
s0:¬full(Y)
ff*e’
Ytills(Y)
e’ cause s
s:full(Y)
s0 ⊃⊂e’
++
�����
HHHHH
rPfis , Y
tills(Y) s:full(Y)
fl
�����
HHHHH
comp
KinokassenfiY
tills(Y)
flr
√full
yfull(y)
vfie’,
fl
movemerge: Introduce an eventuality s with the condition’ s:full(Y) ’(s consists in Y being full).Interpret merge as an (anti)-causative construction: ’e’ cause s’
• An (anti)-causative verbal construction is bi-eventive [Marantz(2005)], and perhypothesis, constructions which project causative relations are always bi-eventive.
property rootsJohnny Depp fullte die Kinokassen (mit Geld) die Fullung der Kinokassen’ J.D. filled the cinemas’ tills (with money)’
(i)rP
�� HHKinokassen
√full
(ii)v
���
HHH
rP
�� HHKinokassen t1
v
��HH√full1 v
vP
*s0
s0:¬full(Y)
ff*e’
Ytills(Y)
e’ cause s
s:full(Y)
s0 ⊃⊂e’
++
�����
HHHHH
rPfis , Y
tills(Y) s:full(Y)
fl
�����
HHHHH
comp
KinokassenfiY
tills(Y)
flr
√full
yfull(y)
vfie’,
fl
movemerge: Introduce an eventuality s with the condition’ s:full(Y) ’(s consists in Y being full).Interpret merge as an (anti)-causative construction: ’e’ cause s’
• An (anti)-causative verbal construction is bi-eventive [Marantz(2005)], and perhypothesis, constructions which project causative relations are always bi-eventive.
The examples above are instances of productive verb formation patterns.
Simple and relational eventive roots denote event-types to be predicated of an eventintroduced by v; (v-modification). This yields mono-eventive structures. No -ung -nouns.
Property-roots and prepositional heads license structure for internal arguments. Theycreate argument slots to be filled in r(oot)Ps which they head and provide the basis forbi-eventive structures.
Sortal roots introduce arguments that must be related to some other argument in theverbal structure. Thus they typically fill argument slots created by prepositional heads.
bi-eventiveJohnny Depp fullte die Kinokassen (mit Geld) Fullung der Kinokassen’J.D. filled the cinemas’ tills’That what J.D. did caused the result state of the tills being full, which is an intrinsicpart of the predication expressed by the bi-eventive structure. This causal relation takesmany forms and need not be in the control of the agent.Let’s call this ’non-control cause’.mono-eventiveJohnny Depp fullte Geld in die Kinokassen * Fullung des Geldes in die K.k.’J.D. poured money into the cinemas’ tills’
Surprise: J.D. has a bucket full of coins and small notes and pours them into tills.The description provided by the mono-eventive structure does not entail that the tills arefull. J.D can stop pouring at any point without thereby making the description invalid.
• Result state conditions of bi-eventive descriptions are an intrinsic part of thedescription.
• Result state conditions of mono-eventive descriptions are inferred.
bi-eventiveJohnny Depp fullte die Kinokassen (mit Geld) Fullung der Kinokassen’J.D. filled the cinemas’ tills’That what J.D. did caused the result state of the tills being full, which is an intrinsicpart of the predication expressed by the bi-eventive structure. This causal relation takesmany forms and need not be in the control of the agent.Let’s call this ’non-control cause’.mono-eventiveJohnny Depp fullte Geld in die Kinokassen * Fullung des Geldes in die K.k.’J.D. poured money into the cinemas’ tills’
Surprise: J.D. has a bucket full of coins and small notes and pours them into tills.The description provided by the mono-eventive structure does not entail that the tills arefull. J.D can stop pouring at any point without thereby making the description invalid.
• Result state conditions of bi-eventive descriptions are an intrinsic part of thedescription.
• Result state conditions of mono-eventive descriptions are inferred.
bi-eventiveJohnny Depp fullte die Kinokassen (mit Geld) Fullung der Kinokassen’J.D. filled the cinemas’ tills’That what J.D. did caused the result state of the tills being full, which is an intrinsicpart of the predication expressed by the bi-eventive structure. This causal relation takesmany forms and need not be in the control of the agent.Let’s call this ’non-control cause’.mono-eventiveJohnny Depp fullte Geld in die Kinokassen * Fullung des Geldes in die K.k.’J.D. poured money into the cinemas’ tills’
Surprise: J.D. has a bucket full of coins and small notes and pours them into tills.The description provided by the mono-eventive structure does not entail that the tills arefull. J.D can stop pouring at any point without thereby making the description invalid.
• Result state conditions of bi-eventive descriptions are an intrinsic part of thedescription.
• Result state conditions of mono-eventive descriptions are inferred.
• mono-eventive verbal descriptions build on unergative syntactic structures
die Kellnerin stellte zwei Glaser hin und fing an, den Tequila einzufullen und fullte,und fullte. (Google)the waitress placed two glasses and started [ filling [in(to)] ] the tequila andpoured and poured
• In this construction√
full does not license an argument slot.
• Still, it is the same root. So some kind of modification must have taken place. Iwill assume that this modification is a case of coercion. The coercion operationcan be seen as one of ’zooming in’.
Assume the following situation: The waitress intends a glass to become full ofTequila. She starts pouring Tequila into it and stops when the glass is full.
The waitress filled a glass (with Tequila) truthfully describes the entire action.The waitress filled and filled truthfully describes any stretch of pouring, betweenwhen she starts and when she stops.
Termination conditions may come from other sources, e.g. through adjunction of aquantized direct object phrase like two deciliters of Tequila.
• mono-eventive verbal descriptions build on unergative syntactic structures
die Kellnerin stellte zwei Glaser hin und fing an, den Tequila einzufullen und fullte,und fullte. (Google)the waitress placed two glasses and started [ filling [in(to)] ] the tequila andpoured and poured
• In this construction√
full does not license an argument slot.
• Still, it is the same root. So some kind of modification must have taken place. Iwill assume that this modification is a case of coercion. The coercion operationcan be seen as one of ’zooming in’.
Assume the following situation: The waitress intends a glass to become full ofTequila. She starts pouring Tequila into it and stops when the glass is full.
The waitress filled a glass (with Tequila) truthfully describes the entire action.The waitress filled and filled truthfully describes any stretch of pouring, betweenwhen she starts and when she stops.
Termination conditions may come from other sources, e.g. through adjunction of aquantized direct object phrase like two deciliters of Tequila.
• mono-eventive verbal descriptions build on unergative syntactic structures
die Kellnerin stellte zwei Glaser hin und fing an, den Tequila einzufullen und fullte,und fullte. (Google)the waitress placed two glasses and started [ filling [in(to)] ] the tequila andpoured and poured
• In this construction√
full does not license an argument slot.
• Still, it is the same root. So some kind of modification must have taken place. Iwill assume that this modification is a case of coercion. The coercion operationcan be seen as one of ’zooming in’.
Assume the following situation: The waitress intends a glass to become full ofTequila. She starts pouring Tequila into it and stops when the glass is full.
The waitress filled a glass (with Tequila) truthfully describes the entire action.The waitress filled and filled truthfully describes any stretch of pouring, betweenwhen she starts and when she stops.
Termination conditions may come from other sources, e.g. through adjunction of aquantized direct object phrase like two deciliters of Tequila.
The semantics of mono-eventive fullen derives from that of bi-eventive fullen: Becauseof its structure, mono-eventive fullen describes (agent-controlled) activities.The properties of these activities are those of the prototypical agentive instances ofbi-eventive fullen. This entails that the activities instantiating mono-eventive fulleninvolve pouring by the agent of fluid or granulated material.
•
bi-eventive mono-eventive
dir. object y qualifying as full;mit-phrase: stuff that y is ’full of’
goal-DP y may qualify as fulldirect object: stuff that y is ’full of’& stuff that can be poured in a li-
teral sense.Das Zimmer mit Rauch fullen(to fill the room with smoke)eine Gans mit Apfeln fullen(to fill the goose with apples)einen Sack mit Apfeln fullento fill a bag with apples
*Rauch in das Zimmer fullen(lit: fill smoke into a the room)* Apfel in die Gans fullen(lit: fill apples into the goose)√
Apfel in einen Sack fullen(lit: fill apples into a bag)
ßung(lit: to ’close” s.th. open)etw. abprcl schließen *Abschließung(lit: to close s.th. so that is is inac-cessible),zuprtcl schließen, *Zuschließung(lit: to close s.th. so that it is shut),
bi-eventive constructions with√schließ
mono-eventive contructions buildingon direct merge with
√schließ
admissible dir. obj.: those thatdenote entities that can be closed
admissible dir. obj.: those that denoteentities that can be closed & can bedirectly manipulated into being clo-sed
den Kreis (the circle), die Augen (theeyes), den Spalt (the fissure) schlie-ßen (to close)
* den Kreis/ *die Augen /*den Spalt*aufschließen*zuschließen*abschließen
die Truhe (the chest), die Tur (thedoor), schließen
die Truhe (the chest), die Tur (thedoor), aufschließen, abschließen, zu-schließen
• What triggers coercion of the property root in direct merge?This is the requirement imposed by the v-head. The v-head introduces aneventuality which is a homogenous process. The property root can act as apredicate of this process if it is re-interpreted as the distinctive property (or set ofproperties) of the prototypical events described by the bi-eventive structure builtfrom the root.
The resulting predication of the process doesn’t entail any culmination.
The mono-eventive structures built using direct merge can often be extended toparticle verb structures. Such particle verbs often have culminations: in these casesit is the particle which contributes the culmination.
• What restricts coercion of the property root in direct merge?Surprise: The set of property roots that undergo coercion is restricted to inherentlyrelational and universal properties.√
full (full) (all parts of the argument have stuff in them);√leer (empty) (all parts of the argument have no stuff);√schließ (close) (all gaps/holes in the argument are blocked);
Engl.√
clean ([Levin(2009)]) (all parts of the argument are free from dirt;)
Note:√
offen (open) is not universal: some gaps/holes in the argument are notblocked.
There are no particle verbs ending on offnen.(In particular: *aufoffnen vs. zuschließen,abschließen).
semantics construction with direct mergeRecall: bi-eventive fullen mono-eventive fullenKinokassen fullen (Fullung d. K.k.) sie fullte und fullte
�����
HHHH
H
rP*s,
Ycs’ tills(Y)s:full(Y)
+
����
HHHH
comp
Kinokassen.fiY Y
tills(Y)
flr
√full
yfull(y)
vPfie’,
fl ?
���
HHH
vfie’,
fl√
full
yfull(y)
⇓
vP*e’,
full(e’)
+
���
HHH
vfie’,
fl√
full
e
full(e)
v requires specifying conditions for the process e’;the root
√full cannot satisfy the requirement per se;
the root full can act as a predicate of e’,if it is reinterpreted as ”manner” in the waydescribed above. (Note: The root’s argument slot y is filtered out).
• The interpretation of direct merge of v with a property root respects the followingrequirements(i) e’ is an intentional action.This is the only option left when vP is not bi-eventive, since the subject is notlicensed by the complement of v.(ii) e’ is atelic.(iii) Direct objects can arise only through adjunction to vP of a prepositional heador particle head. These direct objects must be related to the process introduced byv as participants of the prototypical events described by the bi-eventive structure.
• In a configuration where the property root is the head of a r(oot)P the rootlicences argument structure (the argument slots it introduces into the structure).In mono-eventive structures the root only respects constraints. These are imposedas part of coercion from a property to a ”manner” root.
• The interpretation of direct merge of v with a property root respects the followingrequirements(i) e’ is an intentional action.This is the only option left when vP is not bi-eventive, since the subject is notlicensed by the complement of v.(ii) e’ is atelic.(iii) Direct objects can arise only through adjunction to vP of a prepositional heador particle head. These direct objects must be related to the process introduced byv as participants of the prototypical events described by the bi-eventive structure.
• In a configuration where the property root is the head of a r(oot)P the rootlicences argument structure (the argument slots it introduces into the structure).In mono-eventive structures the root only respects constraints. These are imposedas part of coercion from a property to a ”manner” root.
bi-eventive mono-eventivedir. object satisfies has load(y)DP in mit-phrase: satisfies ’functi-ons as load’(v)
dir. obj. satisfies ’functions asload’(v) & can be manipulated di-rectly
den Wagen mit Heu (be-)laden, (toload the wagon with hay)(die (Be)ladung des Wagens)
Heu auf den Wagen laden(to load hay onto the wagon)(*die Ladung des Heus)
den Kondensator mit Spannung la-den(to charge a condenser with vol-tage)
*Spannung in den Kondensator laden(lit: to load voltage into the conden-ser)
This restriction indicates that the alternates on the right involve mono-eventivestructures obtained via direct merge with a reinterpreted sortal root.
What triggers coercion?A sortal root can act as event predicate provided if it is coerced into one.For instance, the analysis of eine Plane uber den Boden decken (lit: to cover atarpauline over the ground) involves coercing the sortal root
√deck into denoting an
event property that is shared by all prototypical activities leading to their having a cover.A similar focus on prototypical events is involved in the anlysis of graben (to dig).graben has a mono-eventive structure that can be obtained via coercion of a sortal root√
grab to an event predicate. (This sortal root enters as sortal root into the prefix-verbbegraben (to bury s.o. or s.th.) and untergraben (to make a hole under s.th.).Similar effects can be observed for verbal constructions involving roots such as√
Here too the restrictions governing the verbs in question can be explained as the basis ofthe coercion of the root into a prototypical process predicate.
Verbs like (ein Schiff) be-mannen (to man a ship),einen Soldaten besolden, (to pay a soldier),jdm. (sich) kleiden (to dress),do not show such restrictions, thus are presumably built as bi-eventive structures, intowhich the sortal roots
Loose end: direct merge with relational eventive roots and non-monotonicity
I conjecture that absteigen (lit. ’rise down’) is the result of a reinterpretation of the root√steig as a pure event predicate which has lost its argument slot for the internal
subject. This reinterpretation involves the extraction of the manner-like properties ofprototypical agentive instances of steigen, much as we saw in our analysis ofmono-eventive fullen etc.
This is a case study of the contribution of a small selection of roots in ’expected’ and’unexpected’ verbal constructions.The central results are:
• (i) the ’unexpected’ constructions are much more restricted in their applicationsthan the ’expected’ constructions.
(ii) It is the semantics of the roots as it manifests itself in the coerced transitionfrom property or sortal roots to event predicates that is responsible for theserestrictions.
• The explanation of the contribution draws heavily on (a) principles of semanticdecomposition in syntax as proposed by [Marantz(2005)], i.e. bi-eventivity asopposed to mono-eventivity; and (b) the idea of ’direct’ merge, mentioned in[Embick(2004)].
• The syntactic structures built according to these principles allow for the systematiccompositional construction of logically transparent semantic representations.
This is a case study of the contribution of a small selection of roots in ’expected’ and’unexpected’ verbal constructions.The central results are:
• (i) the ’unexpected’ constructions are much more restricted in their applicationsthan the ’expected’ constructions.
(ii) It is the semantics of the roots as it manifests itself in the coerced transitionfrom property or sortal roots to event predicates that is responsible for theserestrictions.
• The explanation of the contribution draws heavily on (a) principles of semanticdecomposition in syntax as proposed by [Marantz(2005)], i.e. bi-eventivity asopposed to mono-eventivity; and (b) the idea of ’direct’ merge, mentioned in[Embick(2004)].
• The syntactic structures built according to these principles allow for the systematiccompositional construction of logically transparent semantic representations.
This is a case study of the contribution of a small selection of roots in ’expected’ and’unexpected’ verbal constructions.The central results are:
• (i) the ’unexpected’ constructions are much more restricted in their applicationsthan the ’expected’ constructions.
(ii) It is the semantics of the roots as it manifests itself in the coerced transitionfrom property or sortal roots to event predicates that is responsible for theserestrictions.
• The explanation of the contribution draws heavily on (a) principles of semanticdecomposition in syntax as proposed by [Marantz(2005)], i.e. bi-eventivity asopposed to mono-eventivity; and (b) the idea of ’direct’ merge, mentioned in[Embick(2004)].
• The syntactic structures built according to these principles allow for the systematiccompositional construction of logically transparent semantic representations.
This is a case study of the contribution of a small selection of roots in ’expected’ and’unexpected’ verbal constructions.The central results are:
• (i) the ’unexpected’ constructions are much more restricted in their applicationsthan the ’expected’ constructions.
(ii) It is the semantics of the roots as it manifests itself in the coerced transitionfrom property or sortal roots to event predicates that is responsible for theserestrictions.
• The explanation of the contribution draws heavily on (a) principles of semanticdecomposition in syntax as proposed by [Marantz(2005)], i.e. bi-eventivity asopposed to mono-eventivity; and (b) the idea of ’direct’ merge, mentioned in[Embick(2004)].
• The syntactic structures built according to these principles allow for the systematiccompositional construction of logically transparent semantic representations.